|Publication number||US4829426 A|
|Application number||US 06/864,162|
|Publication date||May 9, 1989|
|Filing date||May 16, 1986|
|Priority date||Sep 2, 1983|
|Publication number||06864162, 864162, US 4829426 A, US 4829426A, US-A-4829426, US4829426 A, US4829426A|
|Inventors||F. N. Burt|
|Original Assignee||Cogensys Corporation|
|Export Citation||BiBTeX, EndNote, RefMan|
|Patent Citations (2), Referenced by (36), Classifications (5), Legal Events (8)|
|External Links: USPTO, USPTO Assignment, Espacenet|
This application is a continuation of application Ser. No. 529,081, filed Sep. 2, 1983, now U.S. Pat. No. 4,595,982.
A. Field of the Invention
This invention relate to expert systems and methods for decision making using computer-generated models of human decisional processes.
B. Background Art
Many systems have been proposed for use as "expert systems". Among them are Internist by Miller, Pople and Myers; Aesculapius by Shafer, Shafer, Foxlee, and Prost; and AQ11 by Michalski and Larson. These systems are "data driven", that is, they store and make decisions on the basis of values of individual data points. Analysis of these values allows the computer to apply explicitly programmed rules for decisions making. In such programs, the relative importance of the various data points, termed weighting, is arrived at by statistical analysis of the data itself or by the application of probabilistic estimates supplied from human sources. None of these systems has the ability to: "learn" by watching the process of an expert decision-maker; function in more than one problem sphere without extensive reprogramming; emulate the decision-making processes of its teachers; or measure the logical significance of the information it learns to use in dealing with a specific problem.
The present invention involves a computer-based system and method for making decisions within the sphere of a specific problem. An expert mentor provides descriptions of the specific information he uses (data) to make judgments within the sphere of the problem. The expert then makes representative decisions on the computer terminal, characterizing situations in terms of their data values. From this interaction, the computer system of this invention is able to generate an internal representation of the logical processes of the expert and to characterize each situation in terms of a series of intermediate judgments. As a result of structuring, the system of this invention can determine the relative logical importance of each of the data points and to classify each as being of major or minor significance. The system is able to compare the intermediate judgments characterizing any new situation with the stored set of intermediate judgments (the "judgment base") representing its view of the logical sphere of the problem. The logical significance of the new situation can be defined in terms of the major and minor logical differences from the reference situations so stored.
FIG. 1 is a block diagram of system components in the decision-making system.
FIG. 2 is a logic flow diagram of the method of the present invention.
Referring to the drawing, there is shown a block diagram of a decision-making system 10 which includes a processor or CPU 20 of the general purpose type which includes the ability to decode and execute instructions. CPU 20 communicates over bus 13 with memory 14 containing instructions from the program (called the "PARADOCS" program) which controls the operation of system 10. CPU 20 is also in communication over bus 13 with data memory 15 which is used during execution of the instructions stored in memory 14. Processor 20 is also in two-way communication over bus 17 with mass storage device 19 which is used for on-line information storage during program execution and at other times for non-volatile storage.
Processor 20 receives input via bus 22 from input register 24 which receives data from a suitable data input source such as a keyboard, magnetic tape reader, or card register. CPU 20 also transmits output via bus 21 to output register 23 and thence to a suitable data output device 25 which may be a cathode ray tube, printer, or modem.
In general, the PARADOCS program creates a paramorphic model of an expert's decision-making processes by construction of a judgment base. The judgment base contains information about the logical relationships among intermediate judgments used by the expert in his treatment of the problem area. The program is further able to display to users the logical relationships stored in the judgment base.
The PARADOCS program functions within a problem area which is defined by an expert. This expert must act as a mentor to "train" the program in the particular problem area by first giving definition to the data used in representing the problem. This is done by having the mentor list specific pieces of data needed for description of the problem area and providing parameters for each piece of data. These parameters include: (1) the name of the piece of data; (2) the units of measurement for the piece of data; (3) a range of values which has some logical significance for the piece of data; and (4) an indication as to whether entry of the data is always necessary in order to make a decision.
Data may be either scalar or dichotomous, that is, it may have a numeric value or it may be binary. If data is binary it would be either "YES" or "NO". If data is not binary, it is assumed to be scalar and the user will be asked to present a logical range associated with that piece of data. The logical range may be either a normal range, or a significant abnormal range. The mentor will then answer the question "Is this data required--Yes or No?"
The mentor is given the opportunity to define as many pieces of data as he feels are necessary to adequately deal with the problem area. The mentor then defines a reference state by entering a value for each piece of data in the problem area. Finally, the mentor enters an interpretation of what that reference state represents in terms of his professional experience.
The PARADOCS program then generates a "judgment base" against which it will evaluate all future data entered. It generates this judgment base by arriving at an "intermediate judgment" about each of the entered data points. This intermediate judgment is simply the logical relationship between the entered data and the range limits for that data point. That is, whether the data is: (1) Within range; (2) Above range; (3) Below range; (4) Entered without intermediate judgment; (5) Not Entered; (6) Deleted; (7) "No"; (8) "Yes". The judgment base changes dynamically as the PARADOCS program is shown more situations and integrates this experience into the judgment base. The judgment base also contains information about the relative logical significance of each piece of data. PARADOCS analyzes the effect of a change in a given piece of data on the conclusion state generated. It then assigns lesser relative weights to those which have a lesser effect on the totality of the conclusion states. The program then groups these data points and assigns "major" and "minor" classification to the data points.
In adding experiences to the judgment base, PARADOCS accepts only those situations which are logically distinct from all previously encountered situations. The judgment base is, therefore, a logically consistent body of information.
In use, a situation is presented simply by identifying the problem and entering the data representative of that example. The PARADOCS program then executes a series of procedures in order to analyze the significance of the presented situation and report that significance to its user.
PARADOCS examines the judgment base and attempts to determine if the judgment base contains an experience which is logically identical to the one being presented. If a logically identical experience exists in the judgment base, PARADOCS reports the interpretation of that experience as previously given to it by the mentor.
Failing in its attempt to find a logically identical situation in the judgment base, PARADOCS again examines the judgment base in an attempt to locate any experiences which are logically identical to the presented situation only with regard to the data points which are of "major" logical significance. If such experiences are located, PARADOCS reports the interpretation of those experiences along with the minor differences between the presented situation and the stored experience.
The third level of action is similar to the previous one and is used by the program only if no major identical situations are located. It attempts to locate experiences in the judgment base for which major data points match the presented situation to a specified degree. The program then reports the interpretation along with any major and minor differences between the presented situation and the stored experience.
The differences reported by the program can be used by the expert to modify his or her definition of the problem area and thereby refine the expertise of the PARADOCS model. This refinement of the program can take one of several forms: adding more data points to the problem area, deletion of a data point, reinstatement of a previously deleted data point, or "linking" into a new problem area. In any case, the program's performance is paramorphic; it behaves as the expert behaves in his refinement of the problem area.
In the event that the judgment base contains no experiences which are similar to the presented situation, PARADOCS reports this fact and requests that its mentor enter his or her interpretation of the presented situation. This interpretation is stored and the situation is integrated into the judgment base for future reference.
By continually expanding and recalculating the judgment base, the PARADOCS program can become independent of its mentor and operate to provide the benefit of "expertise" beyond even the direct experiences of its mentor to users other than its original mentor.
The PARADOCS program is also able to report to its users the internal structure of its judgment bases. This report is able to display the system's conclusion states and their definitions with regard to the logical interrelationships of the data to the reference state. In addition, the conclusion states which are logically most similar are grouped together.
The program is further able to display the logic of its mentor in a tree-structure and thus communicate the underlying processes of the expert's decisional processes.
The PARADOCS program is composed of a series of interdependent program modules which perform the functions outlined above. These modules are the PROBLEM module, the ANALYSIS module, the REPORT module, the EXAMPLE module, the LISTING Module, the USERFILE module, the TREEPRNT module and the LINKPROB module. Although the modules contained in the appendix are coded in Pascal, it will be understood by one skilled in the art of computer software that any suitable language may be used to implement the functions of the PARADOCS program.
The program modules all create and maintain a series of data files on a mass storage medium of the following structures:
Total Number of Questions
Number Required - Major
Number Not Required - Major
Number Required - Minor
Number Not Required - Minor
Number of Times Used
Date First Entered
Date Last Used
Number of Situations
Minimum Logical Importance for Major Data
Calculate Logical Importance Flag
Primary Problem Code
Calculated Question Flag
Status Code (Active, Deleted, Reinstated)
Short Situation Description
Next Situation Record Number
Problem Link Codes
Date First Entered
Date Last Used
Number of Times Encountered
Next Record Number
Maximum Status Code (Mentor, Teacher, User)
Number of Problems Accessed
Date Put On System
Date Last Used System
Total Number of Times Used
Next Record Number
Status Code (Mentor, Teacher, User)
Number of Times Used
Date of Entry
1st Secondary Key
2nd Secondary Key
3rd Secondary Key
4th Secondary Key
Next Sequence Number
Date of Entry
1st Secondary Key
2nd Secondary Key
3rd Secondary Key
4th Secondary Key
The PROBLEM module is the mentor's method of access to the judgment base. The module first accepts a problem file code from the input device and determines whether that file exists. If the file does not exist, a name for the file is requested and the file is created. The new file will then be filled by entry of the data described above. Alternatively, an existing problem file may be altered in a similar manner. When all data has been entered, the program reevaluates the information about the problem stored in the PROBLEM-FILE at RECORD 000 and re-stores this information. The PROBLEM module then terminates.
The ANALYSIS module generates a situation table, an experience table and a computation table based on the history of the program's experience to date. The module calculates the logical importance (decimal fraction representing the ratio of defined situations which would be undistinguishable if the data point were deleted to the total number of situations), the experiential importance (decimal fraction representing the number of past experiences whose situations would no longer be distinguishable if that data point were deleted to the total number of past experiences for that situation), the logical completeness, the experiential completeness, the realization ratio, the number of possible situations, the number of situations until the model reaches the ability to discriminate, and the number of situations until the model reaches independence of the mentor, for each data point. Further, the PROBLEM-FILE is updated at RECORD 000 and the analysis is displayed to either the system console device or printer.
The EXAMPLE module is the mentor's, users' and teachers' access to the PARADOCS program. For the purpose of entering data descriptive of examples or of new situations, the module is used to obtain an interpretation of a presented situation. The module opens the correct files and accepts or calculates its own answers to each of the questions which define the problem of interest. The system makes intermediate judgments which are coded into an internal representation by formulation of situation strings which are then compared to the SITUATION-FILE as corrected for deletion of reactivation of questions to arrive at a conclusion. The system will also permit linkages to separate problems if required and will transfer all responses stored to date into a separate new judgment base.
The REPORT module generates a REPORT-FILE to the mass storage devices or other output device. The REPORT-FILE contains all information derived from the PARADOCS program in its analysis of a problem. The module generates a summary which includes a summary of the problem, and a description of each situation presented in terms of its differences from a common reference situation and from other logically similar situations.
The LISTING module simply generates a data entry form on the printer for a given problem in case an on-line terminal is unavailable. The questions and allowed answers are taken from the problem stored and formatted into a questionaire.
The USERFILE module allows access to the program's user records. These records control access to the program and define an individual as a mentor, teacher, or user for a given problem. It is this definition which controls the level of interaction allowed for a given operator during collection of data and construction of the paramorphic model.
The TREEPRNT module prints a "decision tree" from the experience of a PARADOCS problem model. This tree is a two dimensional representation of the problem space which graphically demonstrates the logical processes underlying the mentor's expertise.
Lastly, the LINKPROB module allows the definition of a special type of problem which is "linked" to another in a hierarchical manner. A primary problem interpretation may contain a link to one or more secondary problems. Thus, if the primary conclusion is reached, the linked problem may be invoked to further refine the judgment process. Up to 4 levels of linkages are allowed from any conclusion state.
Further information which, when taken with the drawing and claims, will enable those skilled in the art to make and practice the present invention is included herewith as Appendix A and Appendix B.
Appendix A is entitled "PARADOCS Technical Specifications Manual".
Parameters relating to questions, including those constituting type and reference are listed in Appendix A in the description of the SITUATION-FILE at Column 11.
Appendix B comprises the source code listing in the Pascal language of the PARADOCS program in accordance with the present invention. Also included in Appendix B are flowcharts of the procedures used in the program.
While the invention has been shown and described with reference to a specific set of computer instructions, i.e. Pascal, it will be understood that the spirit of this invention can be implemented in other computer languages and can be modified without avoiding the scope of the invention claimed herein.
The Appendices of U.S. Pat. No. 4,595,982 (which is commonly owned by the assignee of the present application) from Column 7, line 15, through Column 527, line 27 are hereby incorporated by reference.
|Cited Patent||Filing date||Publication date||Applicant||Title|
|US4595982 *||Sep 2, 1983||Jun 17, 1986||Burt Frank N||Expert system and method for making decisions in accordance with the decisions of a mentor|
|US4648044 *||Jun 6, 1984||Mar 3, 1987||Teknowledge, Inc.||Basic expert system tool|
|Citing Patent||Filing date||Publication date||Applicant||Title|
|US5159685 *||Dec 6, 1989||Oct 27, 1992||Racal Data Communications Inc.||Expert system for communications network|
|US5224209 *||Apr 23, 1986||Jun 29, 1993||Hitachi, Ltd.||System for choosing between operation modes in a data processing system by interacting with a displayed a multinodal hierarchal figure|
|US5319740 *||Mar 4, 1991||Jun 7, 1994||Hitachi, Ltd.||Expert system building method and system|
|US5337320 *||Oct 28, 1993||Aug 9, 1994||Racal-Datacom, Inc.||Semi-automatic mode of network design|
|US5388189 *||Oct 28, 1993||Feb 7, 1995||Racal-Datacom, Inc.||Alarm filter in an expert system for communications network|
|US5428619 *||Nov 14, 1994||Jun 27, 1995||Digital Equipment Corporation||Model based reasoning system for network fault diagnosis|
|US5557549 *||Mar 28, 1995||Sep 17, 1996||Praxair Technology, Inc.||Knowledge based diagnostic advisory system and method for an air separation plant|
|US5732397 *||Mar 16, 1992||Mar 24, 1998||Lincoln National Risk Management, Inc.||Automated decision-making arrangement|
|US5734890 *||Aug 22, 1995||Mar 31, 1998||Gartner Group||System and method for analyzing procurement decisions and customer satisfaction|
|US5799293 *||Nov 4, 1996||Aug 25, 1998||Ford Global Technologies, Inc.||Method for optimizing the design of a product using knowledge-based engineering techniques|
|US6012051 *||Feb 6, 1997||Jan 4, 2000||America Online, Inc.||Consumer profiling system with analytic decision processor|
|US6529889 *||Jul 27, 1999||Mar 4, 2003||Acappella Software, Inc.||System and method of knowledge architecture|
|US6535863||Apr 6, 2000||Mar 18, 2003||Ford Motor Company||Method for utilizing a knowledge-based system|
|US6598036||Apr 4, 2000||Jul 22, 2003||Ford Global Technologies, Llc||Method for serving engineering rules on a network through servlet and applet|
|US6912515 *||Jun 4, 2001||Jun 28, 2005||Xerox Corporation||Method and system for algorithm synthesis in problem solving|
|US6915268||Jul 27, 2001||Jul 5, 2005||Odyssey Logistics & Technology Corporation||Transport logistics systems and methods|
|US7006980||Jul 6, 2000||Feb 28, 2006||Callvision, Inc.||Method and system for selecting optimal commodities based upon business profile and preferences|
|US7430531||May 30, 2000||Sep 30, 2008||Verisign, Inc.||System and method for assisting customers in choosing a bundled set of commodities using customer preferences|
|US7493277||Aug 21, 2002||Feb 17, 2009||Mydecide Inc.||Business opportunity analytics with dependence|
|US7580878||Dec 6, 2002||Aug 25, 2009||Teradata Us, Inc.||Data fusion for automated business decisions|
|US7707090||Aug 21, 2008||Apr 27, 2010||Globys, Inc.||Method and system for selecting optimal commodities based upon business profile and preferences|
|US7711606||Mar 9, 2005||May 4, 2010||Globys, Inc.||Method and computer readable medium for assisting a customer in choosing among commodities|
|US7774271||Aug 10, 2010||Globys, Inc.||Electronic financial management and analysis system and related methods|
|US7797185||Sep 14, 2010||Mydecide Inc.||Value driven integrated build-to-buy decision analysis system and method|
|US7996310||Aug 9, 2011||Globys, Inc.||Electronic financial management and analysis system and related methods|
|US8019638||Aug 21, 2002||Sep 13, 2011||DecisionStreet, Inc.||Dynamic construction of business analytics|
|US20050091100 *||Dec 7, 2004||Apr 28, 2005||Riggs Glenn E.||Transport logistics systems and methods|
|US20050203793 *||Mar 9, 2005||Sep 15, 2005||Snyder Scott A.||Method and system for selecting optimal commodities based upon business profile and preferences|
|US20060004701 *||Dec 30, 2004||Jan 5, 2006||Bacon Charles F||System and method for adaptive decision making analysis and assessment|
|US20060053027 *||Jul 1, 2005||Mar 9, 2006||Riggs Glenn E||Transport logistics systems and methods|
|US20060265276 *||Jul 26, 2006||Nov 23, 2006||Mydecide, Inc.||Value driven integrated build-to-buy decision analysis system and method|
|US20080040144 *||Oct 17, 2007||Feb 14, 2008||Riggs Glenn E||Transport logistics systems and methods|
|US20100057548 *||Aug 24, 2009||Mar 4, 2010||Globy's,Inc.||Targeted customer offers based on predictive analytics|
|US20110060621 *||Sep 1, 2010||Mar 10, 2011||Mydecide Inc.||Value driven integrated build-to-buy decision analysis system and method|
|WO1997009666A2 *||Aug 26, 1996||Mar 13, 1997||Alex Mintz||Decision making system and method|
|WO1997009666A3 *||Aug 26, 1996||Sep 13, 2001||Alex Mintz||Decision making system and method|
|U.S. Classification||706/45, 706/61|
|Jun 20, 1991||AS||Assignment|
Owner name: CYBERTEK-COGENSYS CORPORATION A CORP. OF DELAWA
Free format text: ASSIGNMENT OF ASSIGNORS INTEREST.;ASSIGNOR:COGENSYS CORPORATION A CORP. OF DELAWARE;REEL/FRAME:005739/0256
Effective date: 19910603
|Jun 15, 1992||FPAY||Fee payment|
Year of fee payment: 4
|Sep 16, 1996||FPAY||Fee payment|
Year of fee payment: 8
|Apr 28, 2000||AS||Assignment|
Owner name: MYND CORPORATION F/K/A CYBERTEK CORPORATION, TEXAS
Free format text: CHANGE OF NAME;ASSIGNOR:CYBERTEK CORPORATION;REEL/FRAME:010795/0007
Effective date: 20000120
Owner name: CYBERTEK CORPORATION, TEXAS
Free format text: MERGER;ASSIGNORS:CYBERTEK-COGENSYS CORPORATION;CYBERTEK CORPORATION;REEL/FRAME:010795/0011
Effective date: 19941227
|May 11, 2000||AS||Assignment|
Owner name: CYBERTEK CORPORATION, TEXAS
Free format text: MERGER;ASSIGNORS:CYBERTEK CORPORATION;CYBERTEK - TEXAS CORPORATION;REEL/FRAME:010795/0171
Effective date: 19911231
|May 18, 2000||AS||Assignment|
Owner name: BANK OF AMERICA, N.A. AS ADMINISTRATIVE AGENT, NOR
Free format text: SECURITY INTEREST;ASSIGNOR:MYND CORPORATION F/K/A CYBERTEK CORPORATION;REEL/FRAME:010841/0177
Effective date: 20000428
|Sep 28, 2000||FPAY||Fee payment|
Year of fee payment: 12
|Mar 12, 2001||AS||Assignment|
Owner name: MYND CORPORATION F/K/A CYBERTEK CORPORATION, SOUTH
Free format text: RELEASE OF SECURITY INTEREST;ASSIGNOR:BANK OF AMERICA, N.A.;REEL/FRAME:011601/0881
Effective date: 20001222